RNAseq

Quality check from Kallisto output

This information is coupled with the multiQC report generated by the pipeline. The percentage of pseudoaligned reads is more homogeneous and higher when the paired-end protocol is used. Looking at all the figures generated, the results seem to be of better quality when using the paired-end protocol.

Protocol

n_targets

n_processed

n_pseudoaligned

n_unique

p_pseudoaligned

p_unique

Protocol + Condition

n_targets

n_processed

n_pseudoaligned

n_unique

p_pseudoaligned

p_unique

Differental Expression : Treated vs Untreated

Single-end

Results :

  • DEGs number : 1931 (padj < 0.05)
  • Up-regulated in Treated condition : 1055
  • Down-regulated in Treated condition : 876
  • Positive logFC = up-regulated in Treated condition / down-regulated in Untreated condition
  • Negative logFC = up-regulated in Untreated condition / down-regulated in Treated condition

Dataset

Liste

Outlier

Heatmap

PCA

Expression

Volcano

Top20

TopHeatmap

MDplot

Paired-end

Results :

  • DEGs number : 3346 (padj < 0.05)
  • Up-regulated in Treated condition : 1773
  • Down-regulated in Treated condition : 1573
  • Positive logFC = up-regulated in Treated condition / down-regulated in Untreated condition
  • Negative logFC = up-regulated in Untreated condition / down-regulated in Treated condition

Dataset

Liste

Outlier

Heatmap

PCA

Expression

Volcano

Top20

TopHeatmap

MDplot

Comparaison Single-end vs Paired-end

Intersection of the DEA result

There are more differentially expressed transcripts detected between the Treated / Untreated conditions using the Paired-end protocol. This seems logical as the data from the paired-end protocol are more homogeneous as shown on the kalisto output concerning the percentage of aligned reads.

What’s more, when the treated protocol is used for PCA, 97% of the treated/non-treated variability is explained, compared with 89% for the single-end protocol. The biological variability that interests us most is much more concentrated than when the single-end protocol is used. As the treatment effect is much more highlighted in the Paried-end protocol, the DEA results will be much more interesting to exploit. The clarity and uniformity of the variability explained by the treatment condition in the paired-end condition may also explain why we obtain a greater number of differentially expressed transcripts compared with the single-end protocol.

The single-end protocol shows that there is more noise in the overall results and is therefore less interesting to use for this type of analysis.

In conclusion, there is a very clear effect of treatment on the individuals sequenced here, whatever the protocol used. It’s just that this effect looks much better highlighted in the data when the paired-end protocol is used.

Multivariate model (EXPERIMENTAL)

Dataset

Outlier

Heatmap

PCA

MDplot

Treated vs Untreated

Results :

  • DEGs number : 4106 (padj < 0.05)
  • Up-regulated in Treated condition : 2034
  • Down-regulated in Treated condition : 2072
  • Positive logFC = up-regulated in Treated condition / down-regulated in Untreated condition
  • Negative logFC = up-regulated in Untreated condition / down-regulated in Treated condition

Expression

Volcano

Top20

TopHeatmap

Single-end vs Paired-end

Results :

  • DEGs number : 5931 (padj < 0.05)
  • Up-regulated in Treated condition : 2457
  • Down-regulated in Treated condition : 3474
  • Positive logFC = up-regulated in Treated condition / down-regulated in Untreated condition
  • Negative logFC = up-regulated in Untreated condition / down-regulated in Treated condition

Expression

Volcano

Top20

TopHeatmap